Tulving and Thompson's (1973) procedure was used to test for both cued recall and recognition of the same set of target words. Analyses indicated that performance on the former was superior. However, if the kind of "syntactical" pairing (noun-adjective, verb -noun or noun-noun) was examined, the effect was restricted only to the noun-adjective pairs. These results suggest a limitation in the generality of Tulving and Thompson's (1973) encoding specificity hypothesis and raise the possibility that their results were artifac tuaI. Tul ving and Thompson (1973) demonstrated that when Ss were induced to encode words in the "context" of another word and then were tested both with a cued recall task and a recognition task, they were able to recall more words than they could recognize. This was so even though the target word (to-be-remembered word) was a weak associate of the cuing word-I% using the Bilodeau and Howell (1965) and Riegel (1965) Field, 1966 ;and Tulving , 1968).Tulving and Thompson (1973) argue th at what is stored about a word on a list is information about the specific encoding of that word in the task context. The stored information mayor may not include the target word's relationship with some other word (or words). If so, then the associated word may be an effective retrievel cue; if not, then it cannot give access to the stored word .The present experiment was prompted by an examination of Tulving and Thompson's (1973) results and preliminary pilot work which replicated their basic results. In both sets of data, it was observed that the word pairs could be classified as noun-adjective (NA), verb-noun (VN) , and noun-noun (NN), and that the difference in recall and recognition may have been related to the noun-adjective pairs . The present study further investigates this finding .